Coarse to fine non-rigid registration: a chain of scale-specific neural networks for multimodal image alignment with application to remote sensing
This addresses the problem of slow and feature-dependent registration for remote sensing and medical imaging, offering a faster and more accurate solution, though it appears incremental as it builds on neural network approaches.
The paper tackles multimodal image non-rigid registration by proposing a chain of scale-specific neural networks that learn features and predict deformations directly, eliminating gradient descent. It demonstrates performance in remote sensing tasks, such as registering cadastral maps and road polylines onto RGB images, and outperforms current keypoint matching methods.
We tackle here the problem of multimodal image non-rigid registration, which is of prime importance in remote sensing and medical imaging. The difficulties encountered by classical registration approaches include feature design and slow optimization by gradient descent. By analyzing these methods, we note the significance of the notion of scale. We design easy-to-train, fully-convolutional neural networks able to learn scale-specific features. Once chained appropriately, they perform global registration in linear time, getting rid of gradient descent schemes by predicting directly the deformation.We show their performance in terms of quality and speed through various tasks of remote sensing multimodal image alignment. In particular, we are able to register correctly cadastral maps of buildings as well as road polylines onto RGB images, and outperform current keypoint matching methods.